AstroSpace v1.0: an exhaustively enumerated CHNO chemical space (≤ 6 heavy atoms) with GFN2-xTB dipole moments and rotational constants for evaluating machine-learning prioritization of interstellar molecules

AstroSpace is the complete set of 35,482 neutral, closed-shell constitutional isomers of carbon, hydrogen, nitrogen and oxygen with one to six heavy atoms, with the GFN2-xTB (tblite) geometry-optimized dipole moment, rotational constants (A, B, C), total energy, rotational partition function at 10 K and rotational line-intensity factor of the 32,458 molecules for which the calculation converged. Each molecule carries a binary label indicating whether it is a confirmed interstellar or circumstellar detection first reported through 2021 (McGuire 2021 census as maintained in the astromol package v2021.7.0): 80 confirmed molecules are in the space, 79 with converged properties, and 8 further in-range census molecules can be written only with formally separated charges and are absent by construction. Eight molecules first reported in 2022–2023 are flagged separately for a prospective check and are not positives. The record also contains the standalone interstellar reference set (96 molecules with names, SMILES, formulae and years of first report), the QM9 B3LYP/6-31G(2df,p) reference dipole moments used for the Δ-ML sensitivity analysis, the candidate tables of the manuscript (Tables S1–S3), a machine-readable file with every number quoted in the manuscript (paper_numbers.json), all figures, and the notebook AstroSpace.ipynb that regenerates everything from scratch on a standard cloud notebook environment (also maintained on GitHub). The data underlie the manuscript AstroSpace: A Size-Controlled, Exhaustively Enumerated CHNO Chemical Space for Evaluating Machine-Learning Prioritization of Interstellar Molecules (Khairbek, Alzahrani, Dessoky, Mahmoud and Thomas, submitted to ChemPlusChem), which uses the space to show that the conventional QM9-based evaluation of machine-learning prioritization is confounded by molecular size and to evaluate physical and learned predictors of detection status under a same-size protocol. See README.md in the record for the data dictionary, provenance and licensing (data CC BY 4.0; code MIT; QM9-derived reference file CC0).